Journal of the Japan Petroleum Institute, 59, (6), 283-292 (2016) 283

[Regular Paper]

Seismic Reservoir Characterization Using Model Based Post-stack Seismic

Inversion: In Case of Fenchuganj Gas Field, Bangladesh

Shefa Ul KARIM1), Md Shoqul ISLAM1), Mohammad Moinul HOSSAIN2), and Md Aminul ISLAM3)
1)
Dept. of Petroleum and Mining Engineering, Shahjalal University of Science and Technology, Sylhet 3114, BANGLADESH
2)
Geophysical Div., Bangladesh Petroleum Exploration and Production Co., Ltd., Dhaka, BANGLADESH
3)
Dept. of Petroleum Geoscience, Faculty of Science, Universiti Brunei Darussalam,
Jalan Tungku Link, Gadong BE1410, NEGARA BRUNEI DARUSSALAM

(Received August 31, 2015)

In the study, model based post-stack inversion technique was used to create pseudo logs at each seismic trace at
the well location to constitute high resolution acoustic inverted impedance models. Interpretation of GR log, SP
log, Caliper log, and seven sand bodies were marked as reservoir zones in well FG #X, which were identied as a
hydrocarbon bearing reservoir. All of predicted gases bearing zones in the well FG #X show the low acoustic
impedance (AI) values in inverted section analogous with the calculated AI value of logging data. The imped-
ance value in an inverted section during 1660-1980 ms represents an image of the alteration of thin sand and a
thick shale layer of Upper Bhubon formation. By observing the relatively lower AI values in the inverted section
three locations have marked as additional well locations (PW 1, PW 2 and PW 3), which are more prospective for
optimizing the gas recovery from this eld.

Keywords
Surma Basin, Wavelet, Reservoir, Seismic trace, Post-stack inversion, Acoustic impedance

1. Introduction
Reservoir characterization plays an essential role in
the prediction of the reservoir properties as well into the
economic potential of the field. A detailed study of
the static behavior of a producing hydrocarbon reser-
voir is essential for the development planning of the
reservoir and to reduce risk and uncertainty in choosing
new drilling locations1),2). Perception of reservoir
characterization requires integrated analysis and under-
standing of the available data, such as seismic data and
well log data.
In any seismic reservoir characterization studies, the
rst step towards a successful hydrocarbon discovery is
the mastering of a good subsurface image of seismic
data. The reected seismic wave amplitudes are func-
tions of acoustic/elastic impedances, which project a
contrast (reflection coefficient) between the lithology
above and below a reflecting boundary1)4). Seismic
data is the output of a convolution operation between
the earth reectivity and a source wavelet. The convo-
lution operation produces a band-limited trace, the
bandwidth of which is determined by the seismic wave-
DOI: dx.doi.org/10.1627/jpi.59.283
To whom correspondence should be addressed.
E-mail: sho_q@yahoo.com
let. Also, due to the band limited nature of the seismic
data, lack of low frequencies prevent the transformed
impedance trace from gaining the basic impedances or
velocity structure, which is crucial to making a geologi-
cal interpretation5). An attempt to recover this resolu-
tion is usually made by obtaining the reflectivity
through a deconvolution operation, which is an inverse
problem6).
Seismic inversion is the process of extracting infor-
mation about elastic rock properties from seismic data
based on the travel-time, amplitude, and phase informa-
tion contained within a seismogram7). It is an optimal
w a y t o g e t a b e t t e r s u b s u r f a c e i m a g e 1),5),8)11).
Commercially, different seismic inversion methods are
used to map the detailed reservoir properties such as
lithology and fluid properties5),12). In this research
work, model based post-stack seismic inversion analy-
sis is used to identify the gas bearing potential zones
and possible well locations for further development of
the Fenchuganj gas eld. Only a small part of full res-
ervoir and modified coordinates were studied due to
data condentiality. The Fenchuganj eld has several
pay zones and stratigraphy of the structure is consists
alternate shale and sandstone in varying proportion13).
A conventional seismic interpretation technique is very
uncertain in predicting rock physics. For this reason,
this study intends to reduce the uncertainty to a delinea-

J. Jpn. Petrol. Inst., Vol. 59, No. 6, 2016

 284
Surma Basin is same as Sylhet Trough (after)15). Sediment thickness
in the map is shown in meter.
Fig. 1Generalized Geological Map of Bangladesh and Adjoining
Area
tion of hydrocarbon bearing distribution prediction.
2. Geology and Stratigraphy
The study area is under the Fenchuganj structure
which is situated in the transition zone between the
central Surma Basin and the folded belt in the east and
is closest to the eastern margin of the central Surma
Basin, which is separated in the north from the
Kailashtila Anticline, in the east from the Harargaj
Anticline, and in the south from Batchia Anticline
(Fig. 1)13). In the Surma Basin (Sylhet Trough), it
appears that the Fenchuganj structure is the third elevated
structure, followed by the Chattak and the Atgram
Anticline14). This structure appears as a reversibly
faulted asymmetrical anticline with a NNE-SSW trend-
ing axis. This fault is wider in the southern region,
which becomes narrower towards north. It observed
that the eastern ank of the anticline is steeply dipping
than the western ank due to reverse fault13),14).
Structural and combination traps of Miocene age
occur along stratigraphic boundaries, in sandstone-lled
channel deposits, and in sandstone beds sealed laterally
by shale-lled channels; these comprise major traps in
the eastern part of the basin. In general, these sedi-
mentary strata have been folded into several large-scale
anticlines that are unfaulted or slight to moderate fault
in the western and central parts of the basin15).
The Surma Basin contains a great thickness of
Tertiary sedimentary strata. This basin contains as
much as 20 km of sediments consisting of deltaic, estu-
J. Jpn. Petrol. Inst., Vol. 59,
arine, shallow-marine sandstones, siltstones, and shales
that contain abundant plant-derived organic materials.
The variety of sedimentary facies of the Surma Basin
indicates a range of depositional environment during
Neogene time16). These strata generally contain about
0.5 to 3 % total organic carbon (TOC) although in places
the content of organic matter may range up to 10.5 %.
Thermal maturation is sufcient to generate natural gas
and liquids throughout much of the area15).
The Fenchuganj gas elds geology is similar to that
of other elds situated in Surma Basin. The stratigra-
phy sediments of Fenchuganj structure consists alter-
nate shale and sandstone in varying proportion of
Oligocene to Recent age17). The stratigraphic succes-
sion of the Fenchuganj gas eld is based on geological
data, seismic data, and well data with brief lithological
description are given in Table 1. The reservoirs have
been founded in the Miocene sediments, which are
mainly composed of alternating gray to dark gray clay,
very fine to medium grained sandstones and potential
source rocks include shales and carbonaceous shales of
Eocene, Oligocene, and Miocene age in the basin cen-
ter and in the synclinal troughs between the fold
trends13),18).
3. Methodology
The seismic inversion method is basically a process
of transforming seismic amplitude value to impedance
value. This is done by deconvolution process which
transforms seismic trace to earth reectivity. Inversion
is the subsurface modeling technique to produce a geo-
logic structure using seismic data as input and well data
as control19). The post-stack seismic inversion methods
use stacked (zero-offset) seismic data to produce images
of the AI in depth or time. AI is one of rock-physics
parameters, which is inuenced by the type of lithology,
porosity, fluid content, depth, pressure and tempera-
ture20).
The fundamental concept of seismic exploration is to
send a short time signal into the earth, which is then re-
ected back from a boundary between two units called
reector. The signals are transmitted through the earth
as an elastic wave and brought back to the receiver sub-
surface information such as geological structure, lithol-
ogy and uid through travel time, reection amplitude
and phase variation20). If we assume that the angle of
incidence is zero and that the layers are flat, the
Zoeppritz equations will simplify to the more manage-
able equation given by Eq. (1).
Z i +1 Z i
ri = 1
Z i +1 + Z i
Where ri is the zero-offset P-wave reection coefcient
at the i-th interface of a stack of n-th layer, and Zii
Vi is the AI impedance of the i-th layer.
No. 6, 2016
 285

Table 1 Lithostratigraphic Succession of the Fenchuganj Gas Field

Age Formation Depth [m] Thickness [m] Lithology

Recent Alluvium 0-30 30 Unconsolidated sand, silt and clay.
Late Pliocene DupiTila 30-298 268 Mostly sandstone and minor clay.
sandstone: brown to light brown and coarse.
clay: dark gray, very soft and sticky.
Middle Pliocene Tipam 298-1150 852 Mostly sandstone and minor clay.
sandstone: light to off white, medium grained.
clay: gray to dark gray, soft to moderately hard.
Miocene Upper Bokabil 1150-1466 316 Shale: gray to bluish gray, soft to moderately hard and
compact, and also laminated.
Middle Bokabil 1466-1766 300 Sandstone and shale alteration.
sandstone: light to clear and medium to ne.
shale: laminated, silty, moderately hard.
Lower Bokabil 1766-2236 470 Mostly shale with minor sandstone.
shale: gray to dark gray, hard and compact.
sandstone: light colored and ne to very ne.
Early Miocene Upper Bhuban 2236-4977 914-2741 Alternation of sandstone and shale, with minor calcareous
(vary) siltstone.
sandstone: light to clear and ne grained.
shale: gray to dark gray, hard and compact.
Data source: Geology Div., Bangladesh Petroleum Exploration and Production Co., Ltd., Dhaka.

Equation (1) can be used as a simplified model for
the reflections found on a stacked seismic section.
Actually the recorded seismic trace is the convolution
of the reflectivity with a band limited seismic wavelet
plus some additive noise21).
St = w t rt + n t
Where St is the seismic trace, wt is the seismic wavelet,
rt is the reectivity integrated from depth to time, de-
noting convolution, and nt is the noise component.
Lindseth22) was one of the rst geophysicists to show
that if we assume that the recorded seismic signal is as
given in Eq. (1), it is possible to iteratively obtain the
acoustic impedance in the next layer, Zi1:
1 + ri
Z i+1 = Z i
1 ri
Applying of Eq. (3) to a seismic trace can effectively
transform the seismic reflection data to P-impedance.
The AI for the rst layer needs to be estimated from a
continuous layer above the target area. In this method,
the impedance for the n-th layer can be calculated as
follows11),21).
1 + ri
Z n = Z1
1 ri
This procedure, however, is not free from problems.
The most severe drawback is that the effect of the band-
limited wavelet is supposed to remove the low frequency
component of the reectivity and treat the trace as a set
of reection coefcients, which means that it can never
be exactly recovered Eq. (2). Another prime concern
is inversion involves removing noise component and
proper scaling of seismic data21). To assure a more
realistic result the low-frequency component missing in
the seismic is added the logging data with stacked seis-
mic data11),22). An update approach to inversion is a
model based inversion in which an initial low frequency
model is modied iteratively to give the best t to seis-
mic data9),21). The main steps in the inversion proce-
dure include the data preparation and data input into the
2
software, calibration by tying well logs to the seismic
data, estimation of the wavelet, generation of a low-
resolution initial model, inversion analysis, and inver-
sion. The specific software used in this research to
perform model based inversion of stacked seismic data
is H-R (Hampson-Russell) STRATA program by CGG
Veritas at BAPEX interpretation laboratory under
Geophysical Division.
3. 1. Well to Seismic Tie
3 In well to seismic tie, a synthetic trace was generated
to correlate with recorded seismic trace. Since the
well logs are in depth domain while seismic data is in
the time domain, a check shot data were applied before
correlation to convert well data into time domain1),23).
This process generates a non-linear TWT (two-way
time) reference in TVD (true vertical depth) that is used
to convert linear depth logs to a linear time domain.
The common practice to compute vertical two-way
4
reection time using the P-wave velocity log data v(z)
is as follows:
Z dZ
t Z (Z ) = t 0 + 2 5
Z0 v( z )
Where, t0 is the best estimate of vertical two-way time
to depth z0, tZ(Z) is the two-way time, a sampled func-
tion of depth z. Steps common to the most good tie
processing are shown in the ow chart (Fig. 2) below
include. The resulting trace is displayed on the same
vertical scale as is the seismic section for direct com-

J. Jpn. Petrol. Inst., Vol. 59, No. 6, 2016

 286
Fig. 2 Flow Chart for Seismic Tie24)
parison. To improve the match with the seismic data,
the synthetic seismic trace can be recomputed using dif-
ferent wavelets and filters. Through a trial-and-error
process, the interpreter determines at what point the
s y n t h e t i c t r a c e b e s t fi t s t h e s e i s m i c d a t a 25).
Variations in the quality of the well log data can have a
major impact on the final synthetic display. To get a
more realistic synthetic trace quality control of data was
done, especially in case of well log data5),25).
3. 2. Wavelet Extraction
A seismic wavelet is nothing but the source signa-
ture, which is required during the inversion process of
seismic data. A good wavelet is the core of inver-
sion26). In frequency domain wavelet extraction con-
sists determining the amplitude spectrum and phase
spectrum. The amplitude spectrum is determined from
the autocorrelation function of the data under the usual
assumption of random (or white) reflectivity11).
The phase spectrum is more difficult to determine.
Wavelet can and do change from trace to trace and as a
function of travel-time. So extraction process should
be determining a large set of wavelets for each seismic
section. A practical and useful solution is to extract a
single average wavelet for the entire section12). The
wavelet extraction can be purely deterministic (using
surface receivers and other means), purely statistical
(from the seismic data alone) or using a well log data
which is done by correlating the log and seismic data23).
The phase spectrum is not reliably calculated by statis-
tical method and need to be supplied as a separate
parameter by the user. In contrast to the statistical
method, wavelet using well log extraction procedure
gives more exact wavelet.
3. 3. Initial Model
Initial model provides the low and high-frequency
components missing from the seismic data, which were
J. Jpn. Petrol. Inst., Vol. 59,
used to reduce the non-uniqueness of the solution.
Low frequency cut-off point, several band-pass filters
were applied to the seismic data to the best estimate of
the missing frequency range8). Use interpolation along
the seismic horizons and between the well locations to
obtain the initial AI model. The spatial interpolation
method used in the H-R software utilizes inverse-
distance weighting and works as follows. Denoting
any attribute (for example, the impedance) at well
number i as Li; the corresponding attribute Lout calculated
at any location near the wells is given by the following
equation11).
Lout = L i Wi 6
Where the weights are:
di2
Wi =
di2
And di is the distance between well #i and the loca-
tion of interest. Power (2) used in this weighting
ensures that weights stay constant and equal 1 in the
vicinity of each well.
3. 4. Model Based Inversion
Model based inversion is an approach to inversion
that avoids the problems of recursive inversion by itera-
tively changing a model to give a least-squares t to the
seismic data8). The basis of all the inversion analysis
is the input data. In the post-stack seismic inversion,
stacked seismic volume and the well logs are including
velocity and density log is required. Model based in-
version is a recent approach to inversion, which is based
on the convolution model Eq. (2). If the noise is
uncorrelated with the seismic signal, we can solve the
reectivity satisfying this equation (Eq. (2)). This is a
non-linear and band-limited equation, which can be
solved iteratively11),27). That solution is gradually
improving the t between synthetic traces and the ob-
served seismic data. The solution attempts to simulta-
neously solve the best-t reectivity and minimize the
differences between the observed and predicted seismic
traces9). In H-R software model based inversion prod-
uct is computed as shown in the ow chart (Fig. 3).
4. Result and Discussion
Using P-wave curve and density log with VSP
(Vertical Seismic Profiling) data for well seismic tie
shows that the log curves and check-shot surveys of
study well were inserted which are corresponding to
well locations. In this paper, log data and check-shot
survey of well FG #X and post-stack seismic section
FGS #A shown in Fig. 4 are used to evaluate the poten-
tial for middle to deep zone in the study area. The
post-stack seismic section illustrates a typical cross sec-
tion of an anticline (Fig. 4).
As we know, the well log is the fundamental method
No. 6, 2016
 287

Table 2Interpreted Lithology of the Fenchuganj Gas Field at Well

FG #X Based on Log Data

Estimated depth in meter Lithology Remark

1300-1350 Shale
1350-1375 Sand Zone A
1375-1655 Shale
1655-1680 Sand Zone B
1680-1693 Shale
1693-1705 Sand Zone C
1705-1815 Shale
1815-1850 Sand Zone D
1850-2020 Shale
2020-2080 Sand Zone E
2080-2148 Shale
2148-2154 Sand Zone F
2154-2206 Shale
2206-2260 Sand Zone G
2260-2511 Shale
2511-2526 Sand
2526-2612 Shale
2612-2627 Sand
Data source: Geology Div., Bangladesh Petroleum Exploration and
Production Co., Ltd., Dhaka.

lower density than the same lithology that surrounds the

Fig. 3Flow Chart for Hampson-Russell (H-R) Software for Model
Based Inversion27)
The scale of legend varies from 1.00 to 1.00. Low amplitude may
be indicating the potential zones and high amplitude in shale
sequence. The flank portion of the seismic section has shown an
alternative of two major reection patterns. The middle right of this
seismic section is seismically blind zone due to a fault.
Fig. 4Zero-offset Seismic Section of Line FGS #A with P-wave
Curve from Well FG #X
to reflect the subsurface condition by using physical
properties of rocks which are the key parameters of the
reservoir characterization. However, a potential
hydrocarbon bearing zone can be identified by having
low to medium value of the gamma ray response28),
very negative spontaneous potential (SP) log reading,
with high resistivity, higher value of neutron porosity,
reservoir, and decrease of sonic velocity response25).
Using GR log, SP log and Caliper log, seven sand bodies
zone (A, B, C, D, E, F and G) marked as reservoir
zones were interpreted with high hydrocarbon bearing
potential (Fig. 6) at well FG #X. The depth of the
estimated zones and their lithological properties based
on log data are shown in Tables 2 and 3. Zones B and
C have relatively thin reservoir sand with average den-
sity about 2.3 (g/cm3); and AI value is almost 21,100-
22,900 (ft/s) (g/cm3). Besides zones D and E are the
thickest potential gas bearing zones with low average
density and AI value (Tables 2 and 3).
The well to seismic tie has long been considered an
art for geophysical interpreters24). The synthetic trace
is created to correlate with the recorded seismic trace
and extract a suitable wavelet for inversion. As we
have mentioned in section 3. 2., the wavelet derives
either directly from the seismic data or is computed with
the aid of available well data. The synthetic seismic
can be compared with the seismic trace physically mea-
sured at the well to improve the picking of seismic hori-
zons and to improve the accuracy and resolution of
formations of interest25). At rst, we estimated a sta-
tistical wavelet, but this wavelet gives a poor correlation
coefficient that is below 0.60. Consequently, a full
wavelet is estimated using well log and seismic data
shown in Fig. 5. This wavelet includes an estimation
of the phase obtained by matching the real and synthetic
phase spectra at the well location. The length of
extracting wavelet is 150 ms with 20 ms taper length
and the phase of extraction is set to be a constant phase.
The synthetic trace using full wavelet is represented in
Fig. 6. The synthetic traces using full wavelet, gives

J. Jpn. Petrol. Inst., Vol. 59, No. 6, 2016

 288

The straight line (intersect frequency loop) illustrates the average phase for each of the frequency components.

Fig. 5Full Wavelet in the Time Domains (left) and Frequency Domains (right), Extracted by Using Well Log and Seismic
Data

In right panel: the left traces represents the synthetic trace from the
well and the mid traces represent the average of original seismic
trace. Deep black horizontal lines illustrate interpreted reservoir
zones based on logging data of well FG #X.
Fig. 6Correlation Window by Using the Full Wavelet
The trace data within the model are the synthetic traces computed
during the well correlation and inserted curve is the P-wave curve at
well location.
Fig. 7This Low Frequency Background Impedance Model (initial
model) is Generated from the Downhole Logs of Well FG
#X for the Seismic Inversion by Time Domain Filtering

a high correlation level with composite trace where the
current correlation coefficient is 0.905. Parameter
analysis window in Fig. 6 shows that the time lag is
1 ms and correlation will be slightly improved by the
suggested time shift. We applied the time shift to get
the best fit match with the composite trace. On the
basis of correlation, a new log called P-wave corrected
log (using the H-R STRATA software) which is used in
the building of the initial mode. However, due to band
limited nature of the seismic data, the lowest and the
highest frequencies were missed. Lower frequencies
are the most critical to rock properties, because it leads
to determining fluid, porosity, and all other reservoir
properties26). To make unique solution low frequency
model is needed to supply the low frequency compo-
nent missing from the seismic trace data in the inver-
sion. The initial model was built by interpolating the
AI from targeted well location. The density and sonic
logs in the well permit calculation of the AI response29).
The converted acoustic impedance logs were ltered via
a 10/15 Hz high-cut lter to create an initial AI model
for the inversion shown in Fig. 7. This is important
because we expect only the low-frequency component
of the model to supply the low frequencies missing
from the stacked seismic data.
The Inversion QC (quality control) analysis per-
formed on selected well location means testing a range
of inversion parameters quickly and comparing differ-
ent parameters before performing the actual inversion 20).
The total inversion was performed using both the QC

J. Jpn. Petrol. Inst., Vol. 59, No. 6, 2016

 289

analysis determined that the default inversion parameters vided clear subsurface image and vertical variations of
are satisfactory, and that STRATA calculates a single formation. The discontinuous color attributes at right
global scale, which optimizes the t between the inver- side of the inversion result indicate presence of a fault
sion traces and the actual logs at the well location. In zone. Interpreted reservoir zones through logging are
case the parameter analysis is satisfactory, the inversion
results were applied to the whole seismic volume. The
inversion analysis result at the well location compared
to the original log at well FG #X shown in Fig. 8.
The inverted impedance was comparable to the imped-
ance from the log in the time between 900 to 2000 ms.
The original well log used in inversion ended at the time
level 900 ms and below 2000 ms in Fig. 8; the inverted
impedance (high spiking curve in rst panel) is quite dif-
ferent from the log impedance (low spiking curve in
first panel). Measured error was estimated between
the original AI and inverted AI result is 4173.83 (ft/s)
(g/cm3). The second panel shows the synthetic traces
(left) calculated from this inversion result compared
with the input seismic traces (right). A visual compar-
ison of real seismic data and inverted synthetic trace in
well location shows a good correlation coefficient of
The first panel of this display shows an overlay of two impedance
0.910788 (Fig. 8). The analysis of the inversion prod- curves: the original impedance (low spiking curve) and the nal in-
uct has proved that the result is consistent with the actual version AI (high spiking curve) at the well FG #X. Time limit from
gas presence observed in the existing well in study area. 900 to 2000 ms with error calculation and the impedance mist was
A cross section of the Model Based Post-stack inver- minimized during the inversion (first panel). In second panel, the
sion result is shown in Fig. 9 by applying the same left wiggle seismic traces are generated from the inversion results
while the real seismic data trace is shown in right and the correlation
parameters throughout the entire volume. The trace coefcient is 0.910788.
data in the model is the synthetic trace computed during
well correlation. Inversion product (Fig. 9) has pro- Fig. 8Inversion QC Analysis Window

PW 1, PW 2 and PW 3 are prospect wells (vertical straight line) and zones marked by circles with the low impedance
value are the predicted potential zones. The acoustic impedance values are represented by a color scale: the impedance
increases from 16,592 to 33,263 (ft/s) (g/cm3) as indicated by the variations of color intensity which changes upward
from light to dark. The traces within the model are synthetic traces. Locations marked by horizontal line upon the
P-wave curve are indicated the near well bore gas distribution zones interpreted by wire log data of well FG #X.
Bafing impedance amplitudes at right portion of this section indicate the fault effected part (fractured zone) of the
structure.

Fig. 9Model Based Post-stack Seismic Inverted Section

J. Jpn. Petrol. Inst., Vol. 59, No. 6, 2016

 290

Table 3 Calculated AI Value and Extracted AI Value at Delineate Reservoir Zone through Log Interpretation

Vp Calculated AI AI impedance from

Depth Time Average density
Lithology Remark (from log data) impedance inverted volume
[m] [ms] [g/cm3]
[ft/s] (ft/s)(g/cm3) (ft/s)(g/cm3)
1350-1375 1074-1095 Sand Zone A 2.27 8446 19172 19480
1375-1655 1095-1287 Shale - 2.39 13123 31363 32619
1655-1680 1287-1306 Sand Zone B 2.3 9194 21146 21935
1680-1693 1306-1320 Shale - 2.4 13393 32143 33075
1693-1705 1320-1331 Sand Zone C 2.325 9842 22882 21687
1705-1815 1331-1405 Shale - 2.4 12467 29920 28922
1815-1850 1405-1432 Sand Zone D 2.25 8202 18454 18856
1850-2020 1432-1531 Shale - 2.394 14017 33556 33152
2020-2080 1531-1576 Sand Zone E 2.247 8773 19713 19176
2080-2148 1576-1632 Shale - 2.38 13939 33174 32860
2148-2154 1632-1636 Sand Zone F 2.314 9352 21640 22243
2154-2206 1636-1687 Shale - 2.405 11826 28441 27519

marked by the horizontal line on the inverted section
above inserted P-wave curve. Lower impedance value
takes place in sand zones, but the amount of lowering
depends upon the fluid content of sand. In hydro-
carbon bearing sand, lowering of impedance will be
high if compared to that of water bearing sand23). All
of predicted gas bearing zones in the log section of well
FG #X show low impedance value in the inverted sec-
tion (Fig. 9) with impedance value from 18,856 to
22,243 (ft/s) (g/cm3). On the other hand, shale zones
show high impedance from 27,519 to 33,152 (ft/s) (g/
cm3). The calculated value of acoustic impedance
through log data at potential reservoir zones are almost
same as that of the extracted from inverted section
(Table 3). This acoustic impedance of rocks usually
varies with different factors like depth, tectonic com-
pression, burial history, inter-granular porosity, fracture
porosity, cementation, and types of uid and their satu-
ration. The gas-filled unconsolidated highly porous
sand at fairly shallow depths are seen in seismic sec-
tions as bright or dim spots (Fig. 9). These ampli-
tudes or reection anomalies were also appearing in Al
inverted or impedance section as zones of low AI
amplitude30).
AI values near 1100 ms and left side of the well loca-
tion are almost 19,000-22,000 (ft/s) (g/cm3). Again at
left side of the well location (Fig. 9) and time about
1250 ms acoustic impedance range is about 21,500-
23,000 (ft/s) (g/cm3). In addition, at times around
1300 ms and 1430 ms, both sides of well, AI value are
approximately 22,500-23,500 (ft/s) (g/cm3) and 18,500-
21,000 (ft/s) (g/cm3), respectively. All of these low
acoustic impedance zones are marked as potential res-
ervoir zones with circles, indicative of channel sands of
the Upper Bhuban formation of the Surma Group. On
the other hand, at times near 1280, 1400 and 1600 ms
of the inverted section with higher impedance values
indicate the overlaying shale layer (Fig. 9 and Table 3).
AI magnitudes at these locations are nearly 27,100-
32,250 (ft/s) (g/cm3). The consecutive variations of
the acoustic impedance values in inverted section at
time 1660-1980 ms represent an image of the thin sand
and thick shale layer alteration of Upper Bhuban forma-
tion. Through the observation of relatively lower im-
pedance values in the inverted section three prospect
well locations PW 1, PW 2 and PW 3 are estimated
(proposed) and marked with vertical black lines. A
fault zone is visible in the right part of the inverted sec-
tion rather than the initial model (Fig. 7).
Further, we found that there was a notable difference
between conventional and inverted seismic section.
The result illustrates that inversion product combined
with rock physics technique analysis could provide a
better image to predict vertical section of a reservoir
character distribution comparable to conventional seis-
mic interpretation techniques (Fig. 10). However, due
to lower resolution convention seismic section unable to
produce such structure as clearly visible in inverted sec-
tion which is almost similar to the log interpretation re-
sults. But in stacked seismic section, seismic attribute
provides an image of the discontinuous formation dis-
tribution (Fig. 10). This promising result is a proof of
concept that seismic inversion can be used as one of the
tools to approach the hydrocarbon or reservoir distribu-
tion prediction.
5. Conclusion
By interpretation of GR log, SP log and Caliper log,
seven sand bodies zone (A, B, C, D, E, F and G) were
marked as reservoir zones in well FG #X which were
identied as hydrocarbon bearing reservoir. The syn-
thetic trace using full wavelet gives a high correlation
level with composite trace where the current correlation
coefficient is 0.905. The inversion analysis result at
the well location compared to the original log at well
FG #X provides a good correlation of 0.910788
between the inversion traces and the original logs. All
of predicted gas bearing zones in the well FG #X show
the low impedance values in inverted section and values

J. Jpn. Petrol. Inst., Vol. 59, No. 6, 2016


The Inverted section provides a better image of the vertical section.
Fig. 10Comparison between Model Based Post-stack Inverted Section (left) and Seismic Amplitude Section from
a Conventional Survey
are about 18,856 to 22,243 (ft/s) (g/cm3). Also, the
calculated AI value from logging data at potential zones
are almost same as that of the extracted value from
inverted section. Zones of low acoustic impedance at
near times 1100, 1250, 1300 and 1430 ms, are marked
(enclosed by circle) as potential zones of sand layer.
On the other hand, at near times 1280, 1400 and
1600 ms are higher impedance zones indicating the
overlaying shale layer. Furthermore, the impedance
value in inverted section during time of 1660-1980 ms
represents an image of the alteration of thin sand and a
thick shale layer of Upper Bhuban formation.
Observing the relatively lower impedance values in the
inverted section estimated three prospect well locations
(PW 1, PW 2 and PW 3), which are more prospective
for optimizing the gas recovery from this eld.
Acknowledgment
The authors are grateful to BAPEX authority, espe-
cially Geological and Geophysical Division for allow-
ing us to collect the necessary data and use their soft-
ware for this work.
References
1) Karbalaali, H., Shadizadeh, S. R., Riahi, M. A., Iranian
Journal of Oil & Gas Science and Technology, 2, (2), 8 (2013).
2) Torres, C., Sen, M., Integrated Approach for the Petrophysical
Interpretation of Post- and Pre-stack 3D Seismic Data, Well
Log and Core Data, Geological Information and Production
Data via Bayesian Stochastic Inversion, Third Annual Report,
J. Jpn. Petrol. Inst., Vol. 59,
291
Institute of Geophysics, Texas (2004).
3) Prskalo, S., Journal of Hungarian Geomathematics, 2, 51
(2007).
4) Yilmaz, O., Seismic data analysis, Society of Exploration
Geophysicists, Tulsa, Oklahoma (2001), pp. 74170-2740.
5) Nawaz, U. S., Acoustic and Elastic Impedance Models of
Gullfaks Field by Post-Stack Seismic Inversion, M. Sc.
Thesis, Petroleum Engineering and Applied Geophysics,
Norwegian University of Science and Technology, 2013.
6) Robinson, E. A., Silvia, M. T.., Digital Signal Processing and
Time Series Analysis, Holden-Day, Inc., San Francisco
(1978), p. 156.
7) Sheriff, R. E., Encyclopedic dictionary of applied geophys-
ics, Society of Exploration Geophysicists, 4th edition, (2002).
8) Russell, B., Hampson, D., A Comparison of Post-stack
Seismic Inversion, 61st Annual International Meeting, SEG,
(1991), pp. 876-878.
9) Lee, K., Yoo, D. G., McMechan, G. A., Hwang, N., Lee, G. H.,
Terr. Atmos. Ocean. Sci., 24, (3), 295 (2013).
10) Pendrel, J., CSEG Recorder, 26, (1), 18 (2001).
11) Swisi, A., Post- and Pre-stack Attribute Analysis and Inversion
of Blackfoot 3D Seismic Data Set, M. Sc. Thesis, Geological
Sciences, University of Saskatchewan, Saskatoon, 2009.
12) Russell, H., Theory of the STRATA Program, Hampson-
Russell, CGG Veritas, (1999).
13) Islam, S. M. A., Islam, M. S., Hossain, M. M., Int. J. Oil Gas
Coal Eng., 2, (3), 36 (2014).
14) Hiller, K., Elahi, M., Structural development and hydrocarbon
entrapment in the Surma Basin, Bangladesh (northwest Indo-
Burman fold belt), Singapore Fifth Offshore Southwest
Conference, Singapore, (1984).
15) U. S. Geological Survey-PetroBangla Cooperative Assessment
of Undiscovered Natural Gas Resources of Bangladesh, U. S.
2208-A: 119, (2001).
16) Evans, P., J. Geological Society of India, 5, 80 (1964).
17) Immam, B., Energy Resources of Bangladesh, University
No. 6, 2016
 292
Grant Commission of Bangladesh, 2nd Edition, (2013).
18) Mannan, M. A., Stratigraphic evolution and geochemistry of
the Neogene Surma Group, Surma Basin, Sylhet, Bangladesh, 25)
PhD Dissertation, Department of Geology, University of Oulu,
2002.
19) Sukmono, S., An Introduction to Seismic Reservoir Analysis, 26)
in Seismic Inversion and AVO Analysis for Reservoir
Characterization, Geophysical Engineering, ITB, Bandung
(2002). 27)
20) Hidayat, L., Prediction of Gas Distribution Using Seismic
Simultaneous Inversion Method, M. Sc. Thesis, Mathematics 28)
and Natural Sciences, University of Indonesia, Jakarta, 2009.
21) Russell, B., Hampson, D., Bankhead, B., An Inversion
Primer, CSEG recorder, 31, No. Special, (2006). 29)
22) Lindseth, R. O., Geophysics, 44, (1), 3 (1979). 30)
23) Dubey, A. K., Reservoir Characterization using AVO and
Seismic Inversion Techniques, 9th Biennial Int. Conf. &
Exposition on Petroleum Geophysics, (2012), p. 205.
24) Munoz, A., Hale, D., Automatically tying well logs to seismic
J. Jpn. Petrol. Inst., Vol. 59,
data, Center for Wave Phenomena, Colorado School of Mines,
Golden, CO80401, USA, p. 253-260.
Student chapter, Universitas Gadjah Mada, Well log analysis
for reservoir characterization, America Association of
Petroleum Geologists, 1st place, write-off competition, (2004).
Jain, C., Effect of Seismic Wavelet Phase on Post Stack
Inversion, 10th Biennial Int. Conf. & Exposition, Kochi
(2013), p. 410.
Russell, H., STRATA guide, Hampson-Russell, CGG Veritas,
(2007).
Asquith, G., Krygowski, D., Basic well log analysis, 2nd ed.,
vol. 16, America Association of Petroleum Geologists, Tulsa,
Oklahoma (2004).
Veeken, P. C. H., Silva M. D., First Break, 22, 47 (2004).
Sorneo, E. O., Acoustic impedance inversion and AVO analy-
sis on the Pliocene shallow gas sands, Rio Bravo field, San
Joaquin Basin, California, M. S. Thesis, Southern Methodist
University, 2007.
No. 6, 2016